1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * super.c 4 * 5 * load/unload driver, mount/dismount volumes 6 * 7 * Copyright (C) 2002, 2004 Oracle. All rights reserved. 8 */ 9 10 #include <linux/module.h> 11 #include <linux/fs.h> 12 #include <linux/types.h> 13 #include <linux/slab.h> 14 #include <linux/highmem.h> 15 #include <linux/init.h> 16 #include <linux/random.h> 17 #include <linux/statfs.h> 18 #include <linux/moduleparam.h> 19 #include <linux/blkdev.h> 20 #include <linux/socket.h> 21 #include <linux/inet.h> 22 #include <linux/parser.h> 23 #include <linux/crc32.h> 24 #include <linux/debugfs.h> 25 #include <linux/mount.h> 26 #include <linux/seq_file.h> 27 #include <linux/quotaops.h> 28 #include <linux/signal.h> 29 30 #define CREATE_TRACE_POINTS 31 #include "ocfs2_trace.h" 32 33 #include <cluster/masklog.h> 34 35 #include "ocfs2.h" 36 37 /* this should be the only file to include a version 1 header */ 38 #include "ocfs1_fs_compat.h" 39 40 #include "alloc.h" 41 #include "aops.h" 42 #include "blockcheck.h" 43 #include "dlmglue.h" 44 #include "export.h" 45 #include "extent_map.h" 46 #include "heartbeat.h" 47 #include "inode.h" 48 #include "journal.h" 49 #include "localalloc.h" 50 #include "namei.h" 51 #include "slot_map.h" 52 #include "super.h" 53 #include "sysfile.h" 54 #include "uptodate.h" 55 #include "xattr.h" 56 #include "quota.h" 57 #include "refcounttree.h" 58 #include "suballoc.h" 59 60 #include "buffer_head_io.h" 61 #include "filecheck.h" 62 63 static struct kmem_cache *ocfs2_inode_cachep; 64 struct kmem_cache *ocfs2_dquot_cachep; 65 struct kmem_cache *ocfs2_qf_chunk_cachep; 66 67 static struct dentry *ocfs2_debugfs_root; 68 69 MODULE_AUTHOR("Oracle"); 70 MODULE_LICENSE("GPL"); 71 MODULE_DESCRIPTION("OCFS2 cluster file system"); 72 73 struct mount_options 74 { 75 unsigned long commit_interval; 76 unsigned long mount_opt; 77 unsigned int atime_quantum; 78 unsigned short slot; 79 int localalloc_opt; 80 unsigned int resv_level; 81 int dir_resv_level; 82 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 83 }; 84 85 static int ocfs2_parse_options(struct super_block *sb, char *options, 86 struct mount_options *mopt, 87 int is_remount); 88 static int ocfs2_check_set_options(struct super_block *sb, 89 struct mount_options *options); 90 static int ocfs2_show_options(struct seq_file *s, struct dentry *root); 91 static void ocfs2_put_super(struct super_block *sb); 92 static int ocfs2_mount_volume(struct super_block *sb); 93 static int ocfs2_remount(struct super_block *sb, int *flags, char *data); 94 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); 95 static int ocfs2_initialize_mem_caches(void); 96 static void ocfs2_free_mem_caches(void); 97 static void ocfs2_delete_osb(struct ocfs2_super *osb); 98 99 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); 100 101 static int ocfs2_sync_fs(struct super_block *sb, int wait); 102 103 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); 104 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); 105 static void ocfs2_release_system_inodes(struct ocfs2_super *osb); 106 static int ocfs2_check_volume(struct ocfs2_super *osb); 107 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 108 struct buffer_head *bh, 109 u32 sectsize, 110 struct ocfs2_blockcheck_stats *stats); 111 static int ocfs2_initialize_super(struct super_block *sb, 112 struct buffer_head *bh, 113 int sector_size, 114 struct ocfs2_blockcheck_stats *stats); 115 static int ocfs2_get_sector(struct super_block *sb, 116 struct buffer_head **bh, 117 int block, 118 int sect_size); 119 static struct inode *ocfs2_alloc_inode(struct super_block *sb); 120 static void ocfs2_free_inode(struct inode *inode); 121 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend); 122 static int ocfs2_enable_quotas(struct ocfs2_super *osb); 123 static void ocfs2_disable_quotas(struct ocfs2_super *osb); 124 125 static struct dquot **ocfs2_get_dquots(struct inode *inode) 126 { 127 return OCFS2_I(inode)->i_dquot; 128 } 129 130 static const struct super_operations ocfs2_sops = { 131 .statfs = ocfs2_statfs, 132 .alloc_inode = ocfs2_alloc_inode, 133 .free_inode = ocfs2_free_inode, 134 .drop_inode = ocfs2_drop_inode, 135 .evict_inode = ocfs2_evict_inode, 136 .sync_fs = ocfs2_sync_fs, 137 .put_super = ocfs2_put_super, 138 .remount_fs = ocfs2_remount, 139 .show_options = ocfs2_show_options, 140 .quota_read = ocfs2_quota_read, 141 .quota_write = ocfs2_quota_write, 142 .get_dquots = ocfs2_get_dquots, 143 }; 144 145 enum { 146 Opt_barrier, 147 Opt_err_panic, 148 Opt_err_ro, 149 Opt_intr, 150 Opt_nointr, 151 Opt_hb_none, 152 Opt_hb_local, 153 Opt_hb_global, 154 Opt_data_ordered, 155 Opt_data_writeback, 156 Opt_atime_quantum, 157 Opt_slot, 158 Opt_commit, 159 Opt_localalloc, 160 Opt_localflocks, 161 Opt_stack, 162 Opt_user_xattr, 163 Opt_nouser_xattr, 164 Opt_inode64, 165 Opt_acl, 166 Opt_noacl, 167 Opt_usrquota, 168 Opt_grpquota, 169 Opt_coherency_buffered, 170 Opt_coherency_full, 171 Opt_resv_level, 172 Opt_dir_resv_level, 173 Opt_journal_async_commit, 174 Opt_err_cont, 175 Opt_err, 176 }; 177 178 static const match_table_t tokens = { 179 {Opt_barrier, "barrier=%u"}, 180 {Opt_err_panic, "errors=panic"}, 181 {Opt_err_ro, "errors=remount-ro"}, 182 {Opt_intr, "intr"}, 183 {Opt_nointr, "nointr"}, 184 {Opt_hb_none, OCFS2_HB_NONE}, 185 {Opt_hb_local, OCFS2_HB_LOCAL}, 186 {Opt_hb_global, OCFS2_HB_GLOBAL}, 187 {Opt_data_ordered, "data=ordered"}, 188 {Opt_data_writeback, "data=writeback"}, 189 {Opt_atime_quantum, "atime_quantum=%u"}, 190 {Opt_slot, "preferred_slot=%u"}, 191 {Opt_commit, "commit=%u"}, 192 {Opt_localalloc, "localalloc=%d"}, 193 {Opt_localflocks, "localflocks"}, 194 {Opt_stack, "cluster_stack=%s"}, 195 {Opt_user_xattr, "user_xattr"}, 196 {Opt_nouser_xattr, "nouser_xattr"}, 197 {Opt_inode64, "inode64"}, 198 {Opt_acl, "acl"}, 199 {Opt_noacl, "noacl"}, 200 {Opt_usrquota, "usrquota"}, 201 {Opt_grpquota, "grpquota"}, 202 {Opt_coherency_buffered, "coherency=buffered"}, 203 {Opt_coherency_full, "coherency=full"}, 204 {Opt_resv_level, "resv_level=%u"}, 205 {Opt_dir_resv_level, "dir_resv_level=%u"}, 206 {Opt_journal_async_commit, "journal_async_commit"}, 207 {Opt_err_cont, "errors=continue"}, 208 {Opt_err, NULL} 209 }; 210 211 #ifdef CONFIG_DEBUG_FS 212 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) 213 { 214 struct ocfs2_cluster_connection *cconn = osb->cconn; 215 struct ocfs2_recovery_map *rm = osb->recovery_map; 216 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan; 217 int i, out = 0; 218 unsigned long flags; 219 220 out += scnprintf(buf + out, len - out, 221 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n", 222 "Device", osb->dev_str, osb->uuid_str, 223 osb->fs_generation, osb->vol_label); 224 225 out += scnprintf(buf + out, len - out, 226 "%10s => State: %d Flags: 0x%lX\n", "Volume", 227 atomic_read(&osb->vol_state), osb->osb_flags); 228 229 out += scnprintf(buf + out, len - out, 230 "%10s => Block: %lu Cluster: %d\n", "Sizes", 231 osb->sb->s_blocksize, osb->s_clustersize); 232 233 out += scnprintf(buf + out, len - out, 234 "%10s => Compat: 0x%X Incompat: 0x%X " 235 "ROcompat: 0x%X\n", 236 "Features", osb->s_feature_compat, 237 osb->s_feature_incompat, osb->s_feature_ro_compat); 238 239 out += scnprintf(buf + out, len - out, 240 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount", 241 osb->s_mount_opt, osb->s_atime_quantum); 242 243 if (cconn) { 244 out += scnprintf(buf + out, len - out, 245 "%10s => Stack: %s Name: %*s " 246 "Version: %d.%d\n", "Cluster", 247 (*osb->osb_cluster_stack == '\0' ? 248 "o2cb" : osb->osb_cluster_stack), 249 cconn->cc_namelen, cconn->cc_name, 250 cconn->cc_version.pv_major, 251 cconn->cc_version.pv_minor); 252 } 253 254 spin_lock_irqsave(&osb->dc_task_lock, flags); 255 out += scnprintf(buf + out, len - out, 256 "%10s => Pid: %d Count: %lu WakeSeq: %lu " 257 "WorkSeq: %lu\n", "DownCnvt", 258 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1), 259 osb->blocked_lock_count, osb->dc_wake_sequence, 260 osb->dc_work_sequence); 261 spin_unlock_irqrestore(&osb->dc_task_lock, flags); 262 263 spin_lock(&osb->osb_lock); 264 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:", 265 "Recovery", 266 (osb->recovery_thread_task ? 267 task_pid_nr(osb->recovery_thread_task) : -1)); 268 if (rm->rm_used == 0) 269 out += scnprintf(buf + out, len - out, " None\n"); 270 else { 271 for (i = 0; i < rm->rm_used; i++) 272 out += scnprintf(buf + out, len - out, " %d", 273 rm->rm_entries[i]); 274 out += scnprintf(buf + out, len - out, "\n"); 275 } 276 spin_unlock(&osb->osb_lock); 277 278 out += scnprintf(buf + out, len - out, 279 "%10s => Pid: %d Interval: %lu\n", "Commit", 280 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1), 281 osb->osb_commit_interval); 282 283 out += scnprintf(buf + out, len - out, 284 "%10s => State: %d TxnId: %lu NumTxns: %d\n", 285 "Journal", osb->journal->j_state, 286 osb->journal->j_trans_id, 287 atomic_read(&osb->journal->j_num_trans)); 288 289 out += scnprintf(buf + out, len - out, 290 "%10s => GlobalAllocs: %d LocalAllocs: %d " 291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n", 292 "Stats", 293 atomic_read(&osb->alloc_stats.bitmap_data), 294 atomic_read(&osb->alloc_stats.local_data), 295 atomic_read(&osb->alloc_stats.bg_allocs), 296 atomic_read(&osb->alloc_stats.moves), 297 atomic_read(&osb->alloc_stats.bg_extends)); 298 299 out += scnprintf(buf + out, len - out, 300 "%10s => State: %u Descriptor: %llu Size: %u bits " 301 "Default: %u bits\n", 302 "LocalAlloc", osb->local_alloc_state, 303 (unsigned long long)osb->la_last_gd, 304 osb->local_alloc_bits, osb->local_alloc_default_bits); 305 306 spin_lock(&osb->osb_lock); 307 out += scnprintf(buf + out, len - out, 308 "%10s => InodeSlot: %d StolenInodes: %d, " 309 "MetaSlot: %d StolenMeta: %d\n", "Steal", 310 osb->s_inode_steal_slot, 311 atomic_read(&osb->s_num_inodes_stolen), 312 osb->s_meta_steal_slot, 313 atomic_read(&osb->s_num_meta_stolen)); 314 spin_unlock(&osb->osb_lock); 315 316 out += scnprintf(buf + out, len - out, "OrphanScan => "); 317 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ", 318 os->os_count, os->os_seqno); 319 out += scnprintf(buf + out, len - out, " Last Scan: "); 320 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) 321 out += scnprintf(buf + out, len - out, "Disabled\n"); 322 else 323 out += scnprintf(buf + out, len - out, "%lu seconds ago\n", 324 (unsigned long)(ktime_get_seconds() - os->os_scantime)); 325 326 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n", 327 "Slots", "Num", "RecoGen"); 328 for (i = 0; i < osb->max_slots; ++i) { 329 out += scnprintf(buf + out, len - out, 330 "%10s %c %3d %10d\n", 331 " ", 332 (i == osb->slot_num ? '*' : ' '), 333 i, osb->slot_recovery_generations[i]); 334 } 335 336 return out; 337 } 338 339 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 340 { 341 struct ocfs2_super *osb = inode->i_private; 342 char *buf = NULL; 343 344 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 345 if (!buf) 346 goto bail; 347 348 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE)); 349 350 file->private_data = buf; 351 352 return 0; 353 bail: 354 return -ENOMEM; 355 } 356 357 static int ocfs2_debug_release(struct inode *inode, struct file *file) 358 { 359 kfree(file->private_data); 360 return 0; 361 } 362 363 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 364 size_t nbytes, loff_t *ppos) 365 { 366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, 367 i_size_read(file->f_mapping->host)); 368 } 369 #else 370 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 371 { 372 return 0; 373 } 374 static int ocfs2_debug_release(struct inode *inode, struct file *file) 375 { 376 return 0; 377 } 378 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 379 size_t nbytes, loff_t *ppos) 380 { 381 return 0; 382 } 383 #endif /* CONFIG_DEBUG_FS */ 384 385 static const struct file_operations ocfs2_osb_debug_fops = { 386 .open = ocfs2_osb_debug_open, 387 .release = ocfs2_debug_release, 388 .read = ocfs2_debug_read, 389 .llseek = generic_file_llseek, 390 }; 391 392 static int ocfs2_sync_fs(struct super_block *sb, int wait) 393 { 394 int status; 395 tid_t target; 396 struct ocfs2_super *osb = OCFS2_SB(sb); 397 398 if (ocfs2_is_hard_readonly(osb)) 399 return -EROFS; 400 401 if (wait) { 402 status = ocfs2_flush_truncate_log(osb); 403 if (status < 0) 404 mlog_errno(status); 405 } else { 406 ocfs2_schedule_truncate_log_flush(osb, 0); 407 } 408 409 if (jbd2_journal_start_commit(osb->journal->j_journal, 410 &target)) { 411 if (wait) 412 jbd2_log_wait_commit(osb->journal->j_journal, 413 target); 414 } 415 return 0; 416 } 417 418 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino) 419 { 420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) 421 && (ino == USER_QUOTA_SYSTEM_INODE 422 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE)) 423 return 0; 424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) 425 && (ino == GROUP_QUOTA_SYSTEM_INODE 426 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE)) 427 return 0; 428 return 1; 429 } 430 431 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) 432 { 433 struct inode *new = NULL; 434 int status = 0; 435 int i; 436 437 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 438 if (IS_ERR(new)) { 439 status = PTR_ERR(new); 440 mlog_errno(status); 441 goto bail; 442 } 443 osb->root_inode = new; 444 445 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 446 if (IS_ERR(new)) { 447 status = PTR_ERR(new); 448 mlog_errno(status); 449 goto bail; 450 } 451 osb->sys_root_inode = new; 452 453 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; 454 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { 455 if (!ocfs2_need_system_inode(osb, i)) 456 continue; 457 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 458 if (!new) { 459 ocfs2_release_system_inodes(osb); 460 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 461 mlog_errno(status); 462 mlog(ML_ERROR, "Unable to load system inode %d, " 463 "possibly corrupt fs?", i); 464 goto bail; 465 } 466 // the array now has one ref, so drop this one 467 iput(new); 468 } 469 470 bail: 471 if (status) 472 mlog_errno(status); 473 return status; 474 } 475 476 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) 477 { 478 struct inode *new = NULL; 479 int status = 0; 480 int i; 481 482 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; 483 i < NUM_SYSTEM_INODES; 484 i++) { 485 if (!ocfs2_need_system_inode(osb, i)) 486 continue; 487 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 488 if (!new) { 489 ocfs2_release_system_inodes(osb); 490 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 491 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", 492 status, i, osb->slot_num); 493 goto bail; 494 } 495 /* the array now has one ref, so drop this one */ 496 iput(new); 497 } 498 499 bail: 500 if (status) 501 mlog_errno(status); 502 return status; 503 } 504 505 static void ocfs2_release_system_inodes(struct ocfs2_super *osb) 506 { 507 int i; 508 struct inode *inode; 509 510 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) { 511 inode = osb->global_system_inodes[i]; 512 if (inode) { 513 iput(inode); 514 osb->global_system_inodes[i] = NULL; 515 } 516 } 517 518 inode = osb->sys_root_inode; 519 if (inode) { 520 iput(inode); 521 osb->sys_root_inode = NULL; 522 } 523 524 inode = osb->root_inode; 525 if (inode) { 526 iput(inode); 527 osb->root_inode = NULL; 528 } 529 530 if (!osb->local_system_inodes) 531 return; 532 533 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) { 534 if (osb->local_system_inodes[i]) { 535 iput(osb->local_system_inodes[i]); 536 osb->local_system_inodes[i] = NULL; 537 } 538 } 539 540 kfree(osb->local_system_inodes); 541 osb->local_system_inodes = NULL; 542 } 543 544 /* We're allocating fs objects, use GFP_NOFS */ 545 static struct inode *ocfs2_alloc_inode(struct super_block *sb) 546 { 547 struct ocfs2_inode_info *oi; 548 549 oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS); 550 if (!oi) 551 return NULL; 552 553 oi->i_sync_tid = 0; 554 oi->i_datasync_tid = 0; 555 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot)); 556 557 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); 558 return &oi->vfs_inode; 559 } 560 561 static void ocfs2_free_inode(struct inode *inode) 562 { 563 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); 564 } 565 566 static unsigned long long ocfs2_max_file_offset(unsigned int bbits, 567 unsigned int cbits) 568 { 569 unsigned int bytes = 1 << cbits; 570 unsigned int trim = bytes; 571 unsigned int bitshift = 32; 572 573 /* 574 * i_size and all block offsets in ocfs2 are always 64 bits 575 * wide. i_clusters is 32 bits, in cluster-sized units. So on 576 * 64 bit platforms, cluster size will be the limiting factor. 577 */ 578 579 #if BITS_PER_LONG == 32 580 BUILD_BUG_ON(sizeof(sector_t) != 8); 581 /* 582 * We might be limited by page cache size. 583 */ 584 if (bytes > PAGE_SIZE) { 585 bytes = PAGE_SIZE; 586 trim = 1; 587 /* 588 * Shift by 31 here so that we don't get larger than 589 * MAX_LFS_FILESIZE 590 */ 591 bitshift = 31; 592 } 593 #endif 594 595 /* 596 * Trim by a whole cluster when we can actually approach the 597 * on-disk limits. Otherwise we can overflow i_clusters when 598 * an extent start is at the max offset. 599 */ 600 return (((unsigned long long)bytes) << bitshift) - trim; 601 } 602 603 static int ocfs2_remount(struct super_block *sb, int *flags, char *data) 604 { 605 int incompat_features; 606 int ret = 0; 607 struct mount_options parsed_options; 608 struct ocfs2_super *osb = OCFS2_SB(sb); 609 u32 tmp; 610 611 sync_filesystem(sb); 612 613 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) || 614 !ocfs2_check_set_options(sb, &parsed_options)) { 615 ret = -EINVAL; 616 goto out; 617 } 618 619 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL | 620 OCFS2_MOUNT_HB_NONE; 621 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) { 622 ret = -EINVAL; 623 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 624 goto out; 625 } 626 627 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != 628 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { 629 ret = -EINVAL; 630 mlog(ML_ERROR, "Cannot change data mode on remount\n"); 631 goto out; 632 } 633 634 /* Probably don't want this on remount; it might 635 * mess with other nodes */ 636 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) && 637 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) { 638 ret = -EINVAL; 639 mlog(ML_ERROR, "Cannot enable inode64 on remount\n"); 640 goto out; 641 } 642 643 /* We're going to/from readonly mode. */ 644 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 645 /* Disable quota accounting before remounting RO */ 646 if (*flags & SB_RDONLY) { 647 ret = ocfs2_susp_quotas(osb, 0); 648 if (ret < 0) 649 goto out; 650 } 651 /* Lock here so the check of HARD_RO and the potential 652 * setting of SOFT_RO is atomic. */ 653 spin_lock(&osb->osb_lock); 654 if (osb->osb_flags & OCFS2_OSB_HARD_RO) { 655 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); 656 ret = -EROFS; 657 goto unlock_osb; 658 } 659 660 if (*flags & SB_RDONLY) { 661 sb->s_flags |= SB_RDONLY; 662 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 663 } else { 664 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { 665 mlog(ML_ERROR, "Cannot remount RDWR " 666 "filesystem due to previous errors.\n"); 667 ret = -EROFS; 668 goto unlock_osb; 669 } 670 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); 671 if (incompat_features) { 672 mlog(ML_ERROR, "Cannot remount RDWR because " 673 "of unsupported optional features " 674 "(%x).\n", incompat_features); 675 ret = -EINVAL; 676 goto unlock_osb; 677 } 678 sb->s_flags &= ~SB_RDONLY; 679 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; 680 } 681 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags); 682 unlock_osb: 683 spin_unlock(&osb->osb_lock); 684 /* Enable quota accounting after remounting RW */ 685 if (!ret && !(*flags & SB_RDONLY)) { 686 if (sb_any_quota_suspended(sb)) 687 ret = ocfs2_susp_quotas(osb, 1); 688 else 689 ret = ocfs2_enable_quotas(osb); 690 if (ret < 0) { 691 /* Return back changes... */ 692 spin_lock(&osb->osb_lock); 693 sb->s_flags |= SB_RDONLY; 694 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 695 spin_unlock(&osb->osb_lock); 696 goto out; 697 } 698 } 699 } 700 701 if (!ret) { 702 /* Only save off the new mount options in case of a successful 703 * remount. */ 704 osb->s_mount_opt = parsed_options.mount_opt; 705 osb->s_atime_quantum = parsed_options.atime_quantum; 706 osb->preferred_slot = parsed_options.slot; 707 if (parsed_options.commit_interval) 708 osb->osb_commit_interval = parsed_options.commit_interval; 709 710 if (!ocfs2_is_hard_readonly(osb)) 711 ocfs2_set_journal_params(osb); 712 713 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 714 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? 715 SB_POSIXACL : 0); 716 } 717 out: 718 return ret; 719 } 720 721 static int ocfs2_sb_probe(struct super_block *sb, 722 struct buffer_head **bh, 723 int *sector_size, 724 struct ocfs2_blockcheck_stats *stats) 725 { 726 int status, tmpstat; 727 struct ocfs1_vol_disk_hdr *hdr; 728 struct ocfs2_dinode *di; 729 int blksize; 730 731 *bh = NULL; 732 733 /* may be > 512 */ 734 *sector_size = bdev_logical_block_size(sb->s_bdev); 735 if (*sector_size > OCFS2_MAX_BLOCKSIZE) { 736 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", 737 *sector_size, OCFS2_MAX_BLOCKSIZE); 738 status = -EINVAL; 739 goto bail; 740 } 741 742 /* Can this really happen? */ 743 if (*sector_size < OCFS2_MIN_BLOCKSIZE) 744 *sector_size = OCFS2_MIN_BLOCKSIZE; 745 746 /* check block zero for old format */ 747 status = ocfs2_get_sector(sb, bh, 0, *sector_size); 748 if (status < 0) { 749 mlog_errno(status); 750 goto bail; 751 } 752 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; 753 if (hdr->major_version == OCFS1_MAJOR_VERSION) { 754 mlog(ML_ERROR, "incompatible version: %u.%u\n", 755 hdr->major_version, hdr->minor_version); 756 status = -EINVAL; 757 } 758 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, 759 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { 760 mlog(ML_ERROR, "incompatible volume signature: %8s\n", 761 hdr->signature); 762 status = -EINVAL; 763 } 764 brelse(*bh); 765 *bh = NULL; 766 if (status < 0) { 767 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " 768 "upgraded before mounting with ocfs v2\n"); 769 goto bail; 770 } 771 772 /* 773 * Now check at magic offset for 512, 1024, 2048, 4096 774 * blocksizes. 4096 is the maximum blocksize because it is 775 * the minimum clustersize. 776 */ 777 status = -EINVAL; 778 for (blksize = *sector_size; 779 blksize <= OCFS2_MAX_BLOCKSIZE; 780 blksize <<= 1) { 781 tmpstat = ocfs2_get_sector(sb, bh, 782 OCFS2_SUPER_BLOCK_BLKNO, 783 blksize); 784 if (tmpstat < 0) { 785 status = tmpstat; 786 mlog_errno(status); 787 break; 788 } 789 di = (struct ocfs2_dinode *) (*bh)->b_data; 790 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); 791 spin_lock_init(&stats->b_lock); 792 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats); 793 if (tmpstat < 0) { 794 brelse(*bh); 795 *bh = NULL; 796 } 797 if (tmpstat != -EAGAIN) { 798 status = tmpstat; 799 break; 800 } 801 } 802 803 bail: 804 return status; 805 } 806 807 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 808 { 809 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL; 810 811 if (osb->s_mount_opt & hb_enabled) { 812 if (ocfs2_mount_local(osb)) { 813 mlog(ML_ERROR, "Cannot heartbeat on a locally " 814 "mounted device.\n"); 815 return -EINVAL; 816 } 817 if (ocfs2_userspace_stack(osb)) { 818 mlog(ML_ERROR, "Userspace stack expected, but " 819 "o2cb heartbeat arguments passed to mount\n"); 820 return -EINVAL; 821 } 822 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) && 823 !ocfs2_cluster_o2cb_global_heartbeat(osb)) || 824 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) && 825 ocfs2_cluster_o2cb_global_heartbeat(osb))) { 826 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n"); 827 return -EINVAL; 828 } 829 } 830 831 if (!(osb->s_mount_opt & hb_enabled)) { 832 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 833 !ocfs2_userspace_stack(osb)) { 834 mlog(ML_ERROR, "Heartbeat has to be started to mount " 835 "a read-write clustered device.\n"); 836 return -EINVAL; 837 } 838 } 839 840 return 0; 841 } 842 843 /* 844 * If we're using a userspace stack, mount should have passed 845 * a name that matches the disk. If not, mount should not 846 * have passed a stack. 847 */ 848 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, 849 struct mount_options *mopt) 850 { 851 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { 852 mlog(ML_ERROR, 853 "cluster stack passed to mount, but this filesystem " 854 "does not support it\n"); 855 return -EINVAL; 856 } 857 858 if (ocfs2_userspace_stack(osb) && 859 strncmp(osb->osb_cluster_stack, mopt->cluster_stack, 860 OCFS2_STACK_LABEL_LEN)) { 861 mlog(ML_ERROR, 862 "cluster stack passed to mount (\"%s\") does not " 863 "match the filesystem (\"%s\")\n", 864 mopt->cluster_stack, 865 osb->osb_cluster_stack); 866 return -EINVAL; 867 } 868 869 return 0; 870 } 871 872 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) 873 { 874 int type; 875 struct super_block *sb = osb->sb; 876 unsigned int feature[OCFS2_MAXQUOTAS] = { 877 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 878 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 879 int status = 0; 880 881 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 882 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 883 continue; 884 if (unsuspend) 885 status = dquot_resume(sb, type); 886 else { 887 struct ocfs2_mem_dqinfo *oinfo; 888 889 /* Cancel periodic syncing before suspending */ 890 oinfo = sb_dqinfo(sb, type)->dqi_priv; 891 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 892 status = dquot_suspend(sb, type); 893 } 894 if (status < 0) 895 break; 896 } 897 if (status < 0) 898 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on " 899 "remount (error = %d).\n", status); 900 return status; 901 } 902 903 static int ocfs2_enable_quotas(struct ocfs2_super *osb) 904 { 905 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL }; 906 struct super_block *sb = osb->sb; 907 unsigned int feature[OCFS2_MAXQUOTAS] = { 908 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 909 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 910 unsigned int ino[OCFS2_MAXQUOTAS] = { 911 LOCAL_USER_QUOTA_SYSTEM_INODE, 912 LOCAL_GROUP_QUOTA_SYSTEM_INODE }; 913 int status; 914 int type; 915 916 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE; 917 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 918 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 919 continue; 920 inode[type] = ocfs2_get_system_file_inode(osb, ino[type], 921 osb->slot_num); 922 if (!inode[type]) { 923 status = -ENOENT; 924 goto out_quota_off; 925 } 926 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2, 927 DQUOT_USAGE_ENABLED); 928 if (status < 0) 929 goto out_quota_off; 930 } 931 932 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 933 iput(inode[type]); 934 return 0; 935 out_quota_off: 936 ocfs2_disable_quotas(osb); 937 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 938 iput(inode[type]); 939 mlog_errno(status); 940 return status; 941 } 942 943 static void ocfs2_disable_quotas(struct ocfs2_super *osb) 944 { 945 int type; 946 struct inode *inode; 947 struct super_block *sb = osb->sb; 948 struct ocfs2_mem_dqinfo *oinfo; 949 950 /* We mostly ignore errors in this function because there's not much 951 * we can do when we see them */ 952 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 953 if (!sb_has_quota_loaded(sb, type)) 954 continue; 955 oinfo = sb_dqinfo(sb, type)->dqi_priv; 956 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 957 inode = igrab(sb->s_dquot.files[type]); 958 /* Turn off quotas. This will remove all dquot structures from 959 * memory and so they will be automatically synced to global 960 * quota files */ 961 dquot_disable(sb, type, DQUOT_USAGE_ENABLED | 962 DQUOT_LIMITS_ENABLED); 963 iput(inode); 964 } 965 } 966 967 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 968 { 969 struct dentry *root; 970 int status, sector_size; 971 struct mount_options parsed_options; 972 struct inode *inode = NULL; 973 struct ocfs2_super *osb = NULL; 974 struct buffer_head *bh = NULL; 975 char nodestr[12]; 976 struct ocfs2_blockcheck_stats stats; 977 978 trace_ocfs2_fill_super(sb, data, silent); 979 980 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 981 status = -EINVAL; 982 goto out; 983 } 984 985 /* probe for superblock */ 986 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); 987 if (status < 0) { 988 mlog(ML_ERROR, "superblock probe failed!\n"); 989 goto out; 990 } 991 992 status = ocfs2_initialize_super(sb, bh, sector_size, &stats); 993 brelse(bh); 994 bh = NULL; 995 if (status < 0) 996 goto out; 997 998 osb = OCFS2_SB(sb); 999 1000 if (!ocfs2_check_set_options(sb, &parsed_options)) { 1001 status = -EINVAL; 1002 goto out_super; 1003 } 1004 osb->s_mount_opt = parsed_options.mount_opt; 1005 osb->s_atime_quantum = parsed_options.atime_quantum; 1006 osb->preferred_slot = parsed_options.slot; 1007 osb->osb_commit_interval = parsed_options.commit_interval; 1008 1009 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt); 1010 osb->osb_resv_level = parsed_options.resv_level; 1011 osb->osb_dir_resv_level = parsed_options.resv_level; 1012 if (parsed_options.dir_resv_level == -1) 1013 osb->osb_dir_resv_level = parsed_options.resv_level; 1014 else 1015 osb->osb_dir_resv_level = parsed_options.dir_resv_level; 1016 1017 status = ocfs2_verify_userspace_stack(osb, &parsed_options); 1018 if (status) 1019 goto out_super; 1020 1021 sb->s_magic = OCFS2_SUPER_MAGIC; 1022 1023 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) | 1024 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0); 1025 1026 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY, 1027 * heartbeat=none */ 1028 if (bdev_read_only(sb->s_bdev)) { 1029 if (!sb_rdonly(sb)) { 1030 status = -EACCES; 1031 mlog(ML_ERROR, "Readonly device detected but readonly " 1032 "mount was not specified.\n"); 1033 goto out_super; 1034 } 1035 1036 /* You should not be able to start a local heartbeat 1037 * on a readonly device. */ 1038 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 1039 status = -EROFS; 1040 mlog(ML_ERROR, "Local heartbeat specified on readonly " 1041 "device.\n"); 1042 goto out_super; 1043 } 1044 1045 status = ocfs2_check_journals_nolocks(osb); 1046 if (status < 0) { 1047 if (status == -EROFS) 1048 mlog(ML_ERROR, "Recovery required on readonly " 1049 "file system, but write access is " 1050 "unavailable.\n"); 1051 goto out_super; 1052 } 1053 1054 ocfs2_set_ro_flag(osb, 1); 1055 1056 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. " 1057 "Cluster services will not be used for this mount. " 1058 "Recovery will be skipped.\n", osb->dev_str); 1059 } 1060 1061 if (!ocfs2_is_hard_readonly(osb)) { 1062 if (sb_rdonly(sb)) 1063 ocfs2_set_ro_flag(osb, 0); 1064 } 1065 1066 status = ocfs2_verify_heartbeat(osb); 1067 if (status < 0) 1068 goto out_super; 1069 1070 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 1071 ocfs2_debugfs_root); 1072 1073 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root, 1074 osb, &ocfs2_osb_debug_fops); 1075 1076 if (ocfs2_meta_ecc(osb)) 1077 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats, 1078 osb->osb_debug_root); 1079 1080 status = ocfs2_mount_volume(sb); 1081 if (status < 0) 1082 goto out_debugfs; 1083 1084 if (osb->root_inode) 1085 inode = igrab(osb->root_inode); 1086 1087 if (!inode) { 1088 status = -EIO; 1089 goto out_dismount; 1090 } 1091 1092 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL, 1093 &ocfs2_kset->kobj); 1094 if (!osb->osb_dev_kset) { 1095 status = -ENOMEM; 1096 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id); 1097 goto out_dismount; 1098 } 1099 1100 /* Create filecheck sysfs related directories/files at 1101 * /sys/fs/ocfs2/<devname>/filecheck */ 1102 if (ocfs2_filecheck_create_sysfs(osb)) { 1103 status = -ENOMEM; 1104 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at " 1105 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id); 1106 goto out_dismount; 1107 } 1108 1109 root = d_make_root(inode); 1110 if (!root) { 1111 status = -ENOMEM; 1112 goto out_dismount; 1113 } 1114 1115 sb->s_root = root; 1116 1117 ocfs2_complete_mount_recovery(osb); 1118 1119 if (ocfs2_mount_local(osb)) 1120 snprintf(nodestr, sizeof(nodestr), "local"); 1121 else 1122 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1123 1124 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1125 "with %s data mode.\n", 1126 osb->dev_str, nodestr, osb->slot_num, 1127 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1128 "ordered"); 1129 1130 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1131 wake_up(&osb->osb_mount_event); 1132 1133 /* Now we can initialize quotas because we can afford to wait 1134 * for cluster locks recovery now. That also means that truncation 1135 * log recovery can happen but that waits for proper quota setup */ 1136 if (!sb_rdonly(sb)) { 1137 status = ocfs2_enable_quotas(osb); 1138 if (status < 0) { 1139 /* We have to err-out specially here because 1140 * s_root is already set */ 1141 mlog_errno(status); 1142 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1143 wake_up(&osb->osb_mount_event); 1144 return status; 1145 } 1146 } 1147 1148 ocfs2_complete_quota_recovery(osb); 1149 1150 /* Now we wake up again for processes waiting for quotas */ 1151 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1152 wake_up(&osb->osb_mount_event); 1153 1154 /* Start this when the mount is almost sure of being successful */ 1155 ocfs2_orphan_scan_start(osb); 1156 1157 return status; 1158 1159 out_dismount: 1160 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1161 wake_up(&osb->osb_mount_event); 1162 ocfs2_dismount_volume(sb, 1); 1163 goto out; 1164 1165 out_debugfs: 1166 debugfs_remove_recursive(osb->osb_debug_root); 1167 out_super: 1168 ocfs2_release_system_inodes(osb); 1169 kfree(osb->recovery_map); 1170 ocfs2_delete_osb(osb); 1171 kfree(osb); 1172 out: 1173 mlog_errno(status); 1174 1175 return status; 1176 } 1177 1178 static struct dentry *ocfs2_mount(struct file_system_type *fs_type, 1179 int flags, 1180 const char *dev_name, 1181 void *data) 1182 { 1183 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super); 1184 } 1185 1186 static struct file_system_type ocfs2_fs_type = { 1187 .owner = THIS_MODULE, 1188 .name = "ocfs2", 1189 .mount = ocfs2_mount, 1190 .kill_sb = kill_block_super, 1191 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1192 .next = NULL 1193 }; 1194 MODULE_ALIAS_FS("ocfs2"); 1195 1196 static int ocfs2_check_set_options(struct super_block *sb, 1197 struct mount_options *options) 1198 { 1199 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA && 1200 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1201 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1202 mlog(ML_ERROR, "User quotas were requested, but this " 1203 "filesystem does not have the feature enabled.\n"); 1204 return 0; 1205 } 1206 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA && 1207 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1208 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1209 mlog(ML_ERROR, "Group quotas were requested, but this " 1210 "filesystem does not have the feature enabled.\n"); 1211 return 0; 1212 } 1213 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL && 1214 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) { 1215 mlog(ML_ERROR, "ACL support requested but extended attributes " 1216 "feature is not enabled\n"); 1217 return 0; 1218 } 1219 /* No ACL setting specified? Use XATTR feature... */ 1220 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL | 1221 OCFS2_MOUNT_NO_POSIX_ACL))) { 1222 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) 1223 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1224 else 1225 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1226 } 1227 return 1; 1228 } 1229 1230 static int ocfs2_parse_options(struct super_block *sb, 1231 char *options, 1232 struct mount_options *mopt, 1233 int is_remount) 1234 { 1235 int status, user_stack = 0; 1236 char *p; 1237 u32 tmp; 1238 int token, option; 1239 substring_t args[MAX_OPT_ARGS]; 1240 1241 trace_ocfs2_parse_options(is_remount, options ? options : "(none)"); 1242 1243 mopt->commit_interval = 0; 1244 mopt->mount_opt = OCFS2_MOUNT_NOINTR; 1245 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1246 mopt->slot = OCFS2_INVALID_SLOT; 1247 mopt->localalloc_opt = -1; 1248 mopt->cluster_stack[0] = '\0'; 1249 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL; 1250 mopt->dir_resv_level = -1; 1251 1252 if (!options) { 1253 status = 1; 1254 goto bail; 1255 } 1256 1257 while ((p = strsep(&options, ",")) != NULL) { 1258 if (!*p) 1259 continue; 1260 1261 token = match_token(p, tokens, args); 1262 switch (token) { 1263 case Opt_hb_local: 1264 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1265 break; 1266 case Opt_hb_none: 1267 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE; 1268 break; 1269 case Opt_hb_global: 1270 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL; 1271 break; 1272 case Opt_barrier: 1273 if (match_int(&args[0], &option)) { 1274 status = 0; 1275 goto bail; 1276 } 1277 if (option) 1278 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1279 else 1280 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1281 break; 1282 case Opt_intr: 1283 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1284 break; 1285 case Opt_nointr: 1286 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1287 break; 1288 case Opt_err_panic: 1289 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1290 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1291 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1292 break; 1293 case Opt_err_ro: 1294 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1295 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1296 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS; 1297 break; 1298 case Opt_err_cont: 1299 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1300 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1301 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT; 1302 break; 1303 case Opt_data_ordered: 1304 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 1305 break; 1306 case Opt_data_writeback: 1307 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 1308 break; 1309 case Opt_user_xattr: 1310 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; 1311 break; 1312 case Opt_nouser_xattr: 1313 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; 1314 break; 1315 case Opt_atime_quantum: 1316 if (match_int(&args[0], &option)) { 1317 status = 0; 1318 goto bail; 1319 } 1320 if (option >= 0) 1321 mopt->atime_quantum = option; 1322 break; 1323 case Opt_slot: 1324 if (match_int(&args[0], &option)) { 1325 status = 0; 1326 goto bail; 1327 } 1328 if (option) 1329 mopt->slot = (u16)option; 1330 break; 1331 case Opt_commit: 1332 if (match_int(&args[0], &option)) { 1333 status = 0; 1334 goto bail; 1335 } 1336 if (option < 0) 1337 return 0; 1338 if (option == 0) 1339 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1340 mopt->commit_interval = HZ * option; 1341 break; 1342 case Opt_localalloc: 1343 if (match_int(&args[0], &option)) { 1344 status = 0; 1345 goto bail; 1346 } 1347 if (option >= 0) 1348 mopt->localalloc_opt = option; 1349 break; 1350 case Opt_localflocks: 1351 /* 1352 * Changing this during remount could race 1353 * flock() requests, or "unbalance" existing 1354 * ones (e.g., a lock is taken in one mode but 1355 * dropped in the other). If users care enough 1356 * to flip locking modes during remount, we 1357 * could add a "local" flag to individual 1358 * flock structures for proper tracking of 1359 * state. 1360 */ 1361 if (!is_remount) 1362 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1363 break; 1364 case Opt_stack: 1365 /* Check both that the option we were passed 1366 * is of the right length and that it is a proper 1367 * string of the right length. 1368 */ 1369 if (((args[0].to - args[0].from) != 1370 OCFS2_STACK_LABEL_LEN) || 1371 (strnlen(args[0].from, 1372 OCFS2_STACK_LABEL_LEN) != 1373 OCFS2_STACK_LABEL_LEN)) { 1374 mlog(ML_ERROR, 1375 "Invalid cluster_stack option\n"); 1376 status = 0; 1377 goto bail; 1378 } 1379 memcpy(mopt->cluster_stack, args[0].from, 1380 OCFS2_STACK_LABEL_LEN); 1381 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1382 /* 1383 * Open code the memcmp here as we don't have 1384 * an osb to pass to 1385 * ocfs2_userspace_stack(). 1386 */ 1387 if (memcmp(mopt->cluster_stack, 1388 OCFS2_CLASSIC_CLUSTER_STACK, 1389 OCFS2_STACK_LABEL_LEN)) 1390 user_stack = 1; 1391 break; 1392 case Opt_inode64: 1393 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1394 break; 1395 case Opt_usrquota: 1396 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1397 break; 1398 case Opt_grpquota: 1399 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1400 break; 1401 case Opt_coherency_buffered: 1402 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED; 1403 break; 1404 case Opt_coherency_full: 1405 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED; 1406 break; 1407 case Opt_acl: 1408 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1409 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; 1410 break; 1411 case Opt_noacl: 1412 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1413 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1414 break; 1415 case Opt_resv_level: 1416 if (is_remount) 1417 break; 1418 if (match_int(&args[0], &option)) { 1419 status = 0; 1420 goto bail; 1421 } 1422 if (option >= OCFS2_MIN_RESV_LEVEL && 1423 option < OCFS2_MAX_RESV_LEVEL) 1424 mopt->resv_level = option; 1425 break; 1426 case Opt_dir_resv_level: 1427 if (is_remount) 1428 break; 1429 if (match_int(&args[0], &option)) { 1430 status = 0; 1431 goto bail; 1432 } 1433 if (option >= OCFS2_MIN_RESV_LEVEL && 1434 option < OCFS2_MAX_RESV_LEVEL) 1435 mopt->dir_resv_level = option; 1436 break; 1437 case Opt_journal_async_commit: 1438 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT; 1439 break; 1440 default: 1441 mlog(ML_ERROR, 1442 "Unrecognized mount option \"%s\" " 1443 "or missing value\n", p); 1444 status = 0; 1445 goto bail; 1446 } 1447 } 1448 1449 if (user_stack == 0) { 1450 /* Ensure only one heartbeat mode */ 1451 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | 1452 OCFS2_MOUNT_HB_GLOBAL | 1453 OCFS2_MOUNT_HB_NONE); 1454 if (hweight32(tmp) != 1) { 1455 mlog(ML_ERROR, "Invalid heartbeat mount options\n"); 1456 status = 0; 1457 goto bail; 1458 } 1459 } 1460 1461 status = 1; 1462 1463 bail: 1464 return status; 1465 } 1466 1467 static int ocfs2_show_options(struct seq_file *s, struct dentry *root) 1468 { 1469 struct ocfs2_super *osb = OCFS2_SB(root->d_sb); 1470 unsigned long opts = osb->s_mount_opt; 1471 unsigned int local_alloc_megs; 1472 1473 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) { 1474 seq_printf(s, ",_netdev"); 1475 if (opts & OCFS2_MOUNT_HB_LOCAL) 1476 seq_printf(s, ",%s", OCFS2_HB_LOCAL); 1477 else 1478 seq_printf(s, ",%s", OCFS2_HB_GLOBAL); 1479 } else 1480 seq_printf(s, ",%s", OCFS2_HB_NONE); 1481 1482 if (opts & OCFS2_MOUNT_NOINTR) 1483 seq_printf(s, ",nointr"); 1484 1485 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1486 seq_printf(s, ",data=writeback"); 1487 else 1488 seq_printf(s, ",data=ordered"); 1489 1490 if (opts & OCFS2_MOUNT_BARRIER) 1491 seq_printf(s, ",barrier=1"); 1492 1493 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1494 seq_printf(s, ",errors=panic"); 1495 else if (opts & OCFS2_MOUNT_ERRORS_CONT) 1496 seq_printf(s, ",errors=continue"); 1497 else 1498 seq_printf(s, ",errors=remount-ro"); 1499 1500 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1501 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1502 1503 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1504 1505 if (osb->osb_commit_interval) 1506 seq_printf(s, ",commit=%u", 1507 (unsigned) (osb->osb_commit_interval / HZ)); 1508 1509 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1510 if (local_alloc_megs != ocfs2_la_default_mb(osb)) 1511 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1512 1513 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1514 seq_printf(s, ",localflocks,"); 1515 1516 if (osb->osb_cluster_stack[0]) 1517 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack, 1518 OCFS2_STACK_LABEL_LEN); 1519 if (opts & OCFS2_MOUNT_USRQUOTA) 1520 seq_printf(s, ",usrquota"); 1521 if (opts & OCFS2_MOUNT_GRPQUOTA) 1522 seq_printf(s, ",grpquota"); 1523 1524 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED) 1525 seq_printf(s, ",coherency=buffered"); 1526 else 1527 seq_printf(s, ",coherency=full"); 1528 1529 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1530 seq_printf(s, ",nouser_xattr"); 1531 else 1532 seq_printf(s, ",user_xattr"); 1533 1534 if (opts & OCFS2_MOUNT_INODE64) 1535 seq_printf(s, ",inode64"); 1536 1537 if (opts & OCFS2_MOUNT_POSIX_ACL) 1538 seq_printf(s, ",acl"); 1539 else 1540 seq_printf(s, ",noacl"); 1541 1542 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL) 1543 seq_printf(s, ",resv_level=%d", osb->osb_resv_level); 1544 1545 if (osb->osb_dir_resv_level != osb->osb_resv_level) 1546 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level); 1547 1548 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 1549 seq_printf(s, ",journal_async_commit"); 1550 1551 return 0; 1552 } 1553 1554 static int __init ocfs2_init(void) 1555 { 1556 int status; 1557 1558 status = init_ocfs2_uptodate_cache(); 1559 if (status < 0) 1560 goto out1; 1561 1562 status = ocfs2_initialize_mem_caches(); 1563 if (status < 0) 1564 goto out2; 1565 1566 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1567 1568 ocfs2_set_locking_protocol(); 1569 1570 status = register_quota_format(&ocfs2_quota_format); 1571 if (status < 0) 1572 goto out3; 1573 status = register_filesystem(&ocfs2_fs_type); 1574 if (!status) 1575 return 0; 1576 1577 unregister_quota_format(&ocfs2_quota_format); 1578 out3: 1579 debugfs_remove(ocfs2_debugfs_root); 1580 ocfs2_free_mem_caches(); 1581 out2: 1582 exit_ocfs2_uptodate_cache(); 1583 out1: 1584 mlog_errno(status); 1585 return status; 1586 } 1587 1588 static void __exit ocfs2_exit(void) 1589 { 1590 unregister_quota_format(&ocfs2_quota_format); 1591 1592 debugfs_remove(ocfs2_debugfs_root); 1593 1594 ocfs2_free_mem_caches(); 1595 1596 unregister_filesystem(&ocfs2_fs_type); 1597 1598 exit_ocfs2_uptodate_cache(); 1599 } 1600 1601 static void ocfs2_put_super(struct super_block *sb) 1602 { 1603 trace_ocfs2_put_super(sb); 1604 1605 ocfs2_sync_blockdev(sb); 1606 ocfs2_dismount_volume(sb, 0); 1607 } 1608 1609 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1610 { 1611 struct ocfs2_super *osb; 1612 u32 numbits, freebits; 1613 int status; 1614 struct ocfs2_dinode *bm_lock; 1615 struct buffer_head *bh = NULL; 1616 struct inode *inode = NULL; 1617 1618 trace_ocfs2_statfs(dentry->d_sb, buf); 1619 1620 osb = OCFS2_SB(dentry->d_sb); 1621 1622 inode = ocfs2_get_system_file_inode(osb, 1623 GLOBAL_BITMAP_SYSTEM_INODE, 1624 OCFS2_INVALID_SLOT); 1625 if (!inode) { 1626 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1627 status = -EIO; 1628 goto bail; 1629 } 1630 1631 status = ocfs2_inode_lock(inode, &bh, 0); 1632 if (status < 0) { 1633 mlog_errno(status); 1634 goto bail; 1635 } 1636 1637 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1638 1639 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1640 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1641 1642 buf->f_type = OCFS2_SUPER_MAGIC; 1643 buf->f_bsize = dentry->d_sb->s_blocksize; 1644 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1645 buf->f_blocks = ((sector_t) numbits) * 1646 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1647 buf->f_bfree = ((sector_t) freebits) * 1648 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1649 buf->f_bavail = buf->f_bfree; 1650 buf->f_files = numbits; 1651 buf->f_ffree = freebits; 1652 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN) 1653 & 0xFFFFFFFFUL; 1654 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN, 1655 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL; 1656 1657 brelse(bh); 1658 1659 ocfs2_inode_unlock(inode, 0); 1660 status = 0; 1661 bail: 1662 iput(inode); 1663 1664 if (status) 1665 mlog_errno(status); 1666 1667 return status; 1668 } 1669 1670 static void ocfs2_inode_init_once(void *data) 1671 { 1672 struct ocfs2_inode_info *oi = data; 1673 1674 oi->ip_flags = 0; 1675 oi->ip_open_count = 0; 1676 spin_lock_init(&oi->ip_lock); 1677 ocfs2_extent_map_init(&oi->vfs_inode); 1678 INIT_LIST_HEAD(&oi->ip_io_markers); 1679 INIT_LIST_HEAD(&oi->ip_unwritten_list); 1680 oi->ip_dir_start_lookup = 0; 1681 init_rwsem(&oi->ip_alloc_sem); 1682 init_rwsem(&oi->ip_xattr_sem); 1683 mutex_init(&oi->ip_io_mutex); 1684 1685 oi->ip_blkno = 0ULL; 1686 oi->ip_clusters = 0; 1687 oi->ip_next_orphan = NULL; 1688 1689 ocfs2_resv_init_once(&oi->ip_la_data_resv); 1690 1691 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1692 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1693 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1694 1695 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), 1696 &ocfs2_inode_caching_ops); 1697 1698 inode_init_once(&oi->vfs_inode); 1699 } 1700 1701 static int ocfs2_initialize_mem_caches(void) 1702 { 1703 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1704 sizeof(struct ocfs2_inode_info), 1705 0, 1706 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1707 SLAB_MEM_SPREAD|SLAB_ACCOUNT), 1708 ocfs2_inode_init_once); 1709 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1710 sizeof(struct ocfs2_dquot), 1711 0, 1712 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1713 SLAB_MEM_SPREAD), 1714 NULL); 1715 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1716 sizeof(struct ocfs2_quota_chunk), 1717 0, 1718 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1719 NULL); 1720 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1721 !ocfs2_qf_chunk_cachep) { 1722 kmem_cache_destroy(ocfs2_inode_cachep); 1723 kmem_cache_destroy(ocfs2_dquot_cachep); 1724 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1725 return -ENOMEM; 1726 } 1727 1728 return 0; 1729 } 1730 1731 static void ocfs2_free_mem_caches(void) 1732 { 1733 /* 1734 * Make sure all delayed rcu free inodes are flushed before we 1735 * destroy cache. 1736 */ 1737 rcu_barrier(); 1738 kmem_cache_destroy(ocfs2_inode_cachep); 1739 ocfs2_inode_cachep = NULL; 1740 1741 kmem_cache_destroy(ocfs2_dquot_cachep); 1742 ocfs2_dquot_cachep = NULL; 1743 1744 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1745 ocfs2_qf_chunk_cachep = NULL; 1746 } 1747 1748 static int ocfs2_get_sector(struct super_block *sb, 1749 struct buffer_head **bh, 1750 int block, 1751 int sect_size) 1752 { 1753 if (!sb_set_blocksize(sb, sect_size)) { 1754 mlog(ML_ERROR, "unable to set blocksize\n"); 1755 return -EIO; 1756 } 1757 1758 *bh = sb_getblk(sb, block); 1759 if (!*bh) { 1760 mlog_errno(-ENOMEM); 1761 return -ENOMEM; 1762 } 1763 lock_buffer(*bh); 1764 if (!buffer_dirty(*bh)) 1765 clear_buffer_uptodate(*bh); 1766 unlock_buffer(*bh); 1767 ll_rw_block(REQ_OP_READ, 1, bh); 1768 wait_on_buffer(*bh); 1769 if (!buffer_uptodate(*bh)) { 1770 mlog_errno(-EIO); 1771 brelse(*bh); 1772 *bh = NULL; 1773 return -EIO; 1774 } 1775 1776 return 0; 1777 } 1778 1779 static int ocfs2_mount_volume(struct super_block *sb) 1780 { 1781 int status = 0; 1782 struct ocfs2_super *osb = OCFS2_SB(sb); 1783 1784 if (ocfs2_is_hard_readonly(osb)) 1785 goto out; 1786 1787 mutex_init(&osb->obs_trim_fs_mutex); 1788 1789 status = ocfs2_dlm_init(osb); 1790 if (status < 0) { 1791 mlog_errno(status); 1792 if (status == -EBADR && ocfs2_userspace_stack(osb)) 1793 mlog(ML_ERROR, "couldn't mount because cluster name on" 1794 " disk does not match the running cluster name.\n"); 1795 goto out; 1796 } 1797 1798 status = ocfs2_super_lock(osb, 1); 1799 if (status < 0) { 1800 mlog_errno(status); 1801 goto out_dlm; 1802 } 1803 1804 /* This will load up the node map and add ourselves to it. */ 1805 status = ocfs2_find_slot(osb); 1806 if (status < 0) { 1807 mlog_errno(status); 1808 goto out_super_lock; 1809 } 1810 1811 /* load all node-local system inodes */ 1812 status = ocfs2_init_local_system_inodes(osb); 1813 if (status < 0) { 1814 mlog_errno(status); 1815 goto out_super_lock; 1816 } 1817 1818 status = ocfs2_check_volume(osb); 1819 if (status < 0) { 1820 mlog_errno(status); 1821 goto out_system_inodes; 1822 } 1823 1824 status = ocfs2_truncate_log_init(osb); 1825 if (status < 0) { 1826 mlog_errno(status); 1827 goto out_system_inodes; 1828 } 1829 1830 ocfs2_super_unlock(osb, 1); 1831 return 0; 1832 1833 out_system_inodes: 1834 if (osb->local_alloc_state == OCFS2_LA_ENABLED) 1835 ocfs2_shutdown_local_alloc(osb); 1836 ocfs2_release_system_inodes(osb); 1837 /* before journal shutdown, we should release slot_info */ 1838 ocfs2_free_slot_info(osb); 1839 ocfs2_journal_shutdown(osb); 1840 out_super_lock: 1841 ocfs2_super_unlock(osb, 1); 1842 out_dlm: 1843 ocfs2_dlm_shutdown(osb, 0); 1844 out: 1845 return status; 1846 } 1847 1848 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1849 { 1850 int tmp, hangup_needed = 0; 1851 struct ocfs2_super *osb = NULL; 1852 char nodestr[12]; 1853 1854 trace_ocfs2_dismount_volume(sb); 1855 1856 BUG_ON(!sb); 1857 osb = OCFS2_SB(sb); 1858 BUG_ON(!osb); 1859 1860 /* Remove file check sysfs related directores/files, 1861 * and wait for the pending file check operations */ 1862 ocfs2_filecheck_remove_sysfs(osb); 1863 1864 kset_unregister(osb->osb_dev_kset); 1865 1866 /* Orphan scan should be stopped as early as possible */ 1867 ocfs2_orphan_scan_stop(osb); 1868 1869 ocfs2_disable_quotas(osb); 1870 1871 /* All dquots should be freed by now */ 1872 WARN_ON(!llist_empty(&osb->dquot_drop_list)); 1873 /* Wait for worker to be done with the work structure in osb */ 1874 cancel_work_sync(&osb->dquot_drop_work); 1875 1876 ocfs2_shutdown_local_alloc(osb); 1877 1878 ocfs2_truncate_log_shutdown(osb); 1879 1880 /* This will disable recovery and flush any recovery work. */ 1881 ocfs2_recovery_exit(osb); 1882 1883 ocfs2_sync_blockdev(sb); 1884 1885 ocfs2_purge_refcount_trees(osb); 1886 1887 /* No cluster connection means we've failed during mount, so skip 1888 * all the steps which depended on that to complete. */ 1889 if (osb->cconn) { 1890 tmp = ocfs2_super_lock(osb, 1); 1891 if (tmp < 0) { 1892 mlog_errno(tmp); 1893 return; 1894 } 1895 } 1896 1897 if (osb->slot_num != OCFS2_INVALID_SLOT) 1898 ocfs2_put_slot(osb); 1899 1900 if (osb->cconn) 1901 ocfs2_super_unlock(osb, 1); 1902 1903 ocfs2_release_system_inodes(osb); 1904 1905 ocfs2_journal_shutdown(osb); 1906 1907 /* 1908 * If we're dismounting due to mount error, mount.ocfs2 will clean 1909 * up heartbeat. If we're a local mount, there is no heartbeat. 1910 * If we failed before we got a uuid_str yet, we can't stop 1911 * heartbeat. Otherwise, do it. 1912 */ 1913 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str && 1914 !ocfs2_is_hard_readonly(osb)) 1915 hangup_needed = 1; 1916 1917 ocfs2_dlm_shutdown(osb, hangup_needed); 1918 1919 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1920 debugfs_remove_recursive(osb->osb_debug_root); 1921 1922 if (hangup_needed) 1923 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1924 1925 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1926 1927 if (ocfs2_mount_local(osb)) 1928 snprintf(nodestr, sizeof(nodestr), "local"); 1929 else 1930 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1931 1932 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1933 osb->dev_str, nodestr); 1934 1935 ocfs2_delete_osb(osb); 1936 kfree(osb); 1937 sb->s_dev = 0; 1938 sb->s_fs_info = NULL; 1939 } 1940 1941 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1942 unsigned uuid_bytes) 1943 { 1944 int i, ret; 1945 char *ptr; 1946 1947 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1948 1949 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1950 if (osb->uuid_str == NULL) 1951 return -ENOMEM; 1952 1953 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1954 /* print with null */ 1955 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1956 if (ret != 2) /* drop super cleans up */ 1957 return -EINVAL; 1958 /* then only advance past the last char */ 1959 ptr += 2; 1960 } 1961 1962 return 0; 1963 } 1964 1965 /* Make sure entire volume is addressable by our journal. Requires 1966 osb_clusters_at_boot to be valid and for the journal to have been 1967 initialized by ocfs2_journal_init(). */ 1968 static int ocfs2_journal_addressable(struct ocfs2_super *osb) 1969 { 1970 int status = 0; 1971 u64 max_block = 1972 ocfs2_clusters_to_blocks(osb->sb, 1973 osb->osb_clusters_at_boot) - 1; 1974 1975 /* 32-bit block number is always OK. */ 1976 if (max_block <= (u32)~0ULL) 1977 goto out; 1978 1979 /* Volume is "huge", so see if our journal is new enough to 1980 support it. */ 1981 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb, 1982 OCFS2_FEATURE_COMPAT_JBD2_SB) && 1983 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0, 1984 JBD2_FEATURE_INCOMPAT_64BIT))) { 1985 mlog(ML_ERROR, "The journal cannot address the entire volume. " 1986 "Enable the 'block64' journal option with tunefs.ocfs2"); 1987 status = -EFBIG; 1988 goto out; 1989 } 1990 1991 out: 1992 return status; 1993 } 1994 1995 static int ocfs2_initialize_super(struct super_block *sb, 1996 struct buffer_head *bh, 1997 int sector_size, 1998 struct ocfs2_blockcheck_stats *stats) 1999 { 2000 int status; 2001 int i, cbits, bbits; 2002 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 2003 struct inode *inode = NULL; 2004 struct ocfs2_super *osb; 2005 u64 total_blocks; 2006 2007 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 2008 if (!osb) { 2009 status = -ENOMEM; 2010 mlog_errno(status); 2011 goto out; 2012 } 2013 2014 sb->s_fs_info = osb; 2015 sb->s_op = &ocfs2_sops; 2016 sb->s_d_op = &ocfs2_dentry_ops; 2017 sb->s_export_op = &ocfs2_export_ops; 2018 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2019 sb->dq_op = &ocfs2_quota_operations; 2020 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP; 2021 sb->s_xattr = ocfs2_xattr_handlers; 2022 sb->s_time_gran = 1; 2023 sb->s_flags |= SB_NOATIME; 2024 /* this is needed to support O_LARGEFILE */ 2025 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2026 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 2027 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 2028 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid, 2029 sizeof(di->id2.i_super.s_uuid)); 2030 2031 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 2032 2033 for (i = 0; i < 3; i++) 2034 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 2035 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2036 2037 osb->sb = sb; 2038 osb->s_sectsize_bits = blksize_bits(sector_size); 2039 BUG_ON(!osb->s_sectsize_bits); 2040 2041 spin_lock_init(&osb->dc_task_lock); 2042 init_waitqueue_head(&osb->dc_event); 2043 osb->dc_work_sequence = 0; 2044 osb->dc_wake_sequence = 0; 2045 INIT_LIST_HEAD(&osb->blocked_lock_list); 2046 osb->blocked_lock_count = 0; 2047 spin_lock_init(&osb->osb_lock); 2048 spin_lock_init(&osb->osb_xattr_lock); 2049 ocfs2_init_steal_slots(osb); 2050 2051 mutex_init(&osb->system_file_mutex); 2052 2053 atomic_set(&osb->alloc_stats.moves, 0); 2054 atomic_set(&osb->alloc_stats.local_data, 0); 2055 atomic_set(&osb->alloc_stats.bitmap_data, 0); 2056 atomic_set(&osb->alloc_stats.bg_allocs, 0); 2057 atomic_set(&osb->alloc_stats.bg_extends, 0); 2058 2059 /* Copy the blockcheck stats from the superblock probe */ 2060 osb->osb_ecc_stats = *stats; 2061 2062 ocfs2_init_node_maps(osb); 2063 2064 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2065 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2066 2067 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2068 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2069 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2070 osb->max_slots); 2071 status = -EINVAL; 2072 goto out; 2073 } 2074 2075 ocfs2_orphan_scan_init(osb); 2076 2077 status = ocfs2_recovery_init(osb); 2078 if (status) { 2079 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 2080 mlog_errno(status); 2081 goto out; 2082 } 2083 2084 init_waitqueue_head(&osb->checkpoint_event); 2085 2086 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 2087 2088 osb->slot_num = OCFS2_INVALID_SLOT; 2089 2090 osb->s_xattr_inline_size = le16_to_cpu( 2091 di->id2.i_super.s_xattr_inline_size); 2092 2093 osb->local_alloc_state = OCFS2_LA_UNUSED; 2094 osb->local_alloc_bh = NULL; 2095 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2096 2097 init_waitqueue_head(&osb->osb_mount_event); 2098 2099 ocfs2_resmap_init(osb, &osb->osb_la_resmap); 2100 2101 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2102 if (!osb->vol_label) { 2103 mlog(ML_ERROR, "unable to alloc vol label\n"); 2104 status = -ENOMEM; 2105 goto out_recovery_map; 2106 } 2107 2108 osb->slot_recovery_generations = 2109 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2110 GFP_KERNEL); 2111 if (!osb->slot_recovery_generations) { 2112 status = -ENOMEM; 2113 mlog_errno(status); 2114 goto out_vol_label; 2115 } 2116 2117 init_waitqueue_head(&osb->osb_wipe_event); 2118 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2119 sizeof(*osb->osb_orphan_wipes), 2120 GFP_KERNEL); 2121 if (!osb->osb_orphan_wipes) { 2122 status = -ENOMEM; 2123 mlog_errno(status); 2124 goto out_slot_recovery_gen; 2125 } 2126 2127 osb->osb_rf_lock_tree = RB_ROOT; 2128 2129 osb->s_feature_compat = 2130 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2131 osb->s_feature_ro_compat = 2132 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2133 osb->s_feature_incompat = 2134 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2135 2136 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2137 mlog(ML_ERROR, "couldn't mount because of unsupported " 2138 "optional features (%x).\n", i); 2139 status = -EINVAL; 2140 goto out_orphan_wipes; 2141 } 2142 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2143 mlog(ML_ERROR, "couldn't mount RDWR because of " 2144 "unsupported optional features (%x).\n", i); 2145 status = -EINVAL; 2146 goto out_orphan_wipes; 2147 } 2148 2149 if (ocfs2_clusterinfo_valid(osb)) { 2150 /* 2151 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null 2152 * terminated, so make sure no overflow happens here by using 2153 * memcpy. Destination strings will always be null terminated 2154 * because osb is allocated using kzalloc. 2155 */ 2156 osb->osb_stackflags = 2157 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags; 2158 memcpy(osb->osb_cluster_stack, 2159 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2160 OCFS2_STACK_LABEL_LEN); 2161 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2162 mlog(ML_ERROR, 2163 "couldn't mount because of an invalid " 2164 "cluster stack label (%s) \n", 2165 osb->osb_cluster_stack); 2166 status = -EINVAL; 2167 goto out_orphan_wipes; 2168 } 2169 memcpy(osb->osb_cluster_name, 2170 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster, 2171 OCFS2_CLUSTER_NAME_LEN); 2172 } else { 2173 /* The empty string is identical with classic tools that 2174 * don't know about s_cluster_info. */ 2175 osb->osb_cluster_stack[0] = '\0'; 2176 } 2177 2178 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2179 2180 /* 2181 * FIXME 2182 * This should be done in ocfs2_journal_init(), but any inode 2183 * writes back operation will cause the filesystem to crash. 2184 */ 2185 status = ocfs2_journal_alloc(osb); 2186 if (status < 0) 2187 goto out_orphan_wipes; 2188 2189 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); 2190 init_llist_head(&osb->dquot_drop_list); 2191 2192 /* get some pseudo constants for clustersize bits */ 2193 osb->s_clustersize_bits = 2194 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2195 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2196 2197 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2198 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2199 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2200 osb->s_clustersize); 2201 status = -EINVAL; 2202 goto out_journal; 2203 } 2204 2205 total_blocks = ocfs2_clusters_to_blocks(osb->sb, 2206 le32_to_cpu(di->i_clusters)); 2207 2208 status = generic_check_addressable(osb->sb->s_blocksize_bits, 2209 total_blocks); 2210 if (status) { 2211 mlog(ML_ERROR, "Volume too large " 2212 "to mount safely on this system"); 2213 status = -EFBIG; 2214 goto out_journal; 2215 } 2216 2217 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2218 sizeof(di->id2.i_super.s_uuid))) { 2219 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2220 status = -ENOMEM; 2221 goto out_journal; 2222 } 2223 2224 strlcpy(osb->vol_label, di->id2.i_super.s_label, 2225 OCFS2_MAX_VOL_LABEL_LEN); 2226 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2227 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2228 osb->first_cluster_group_blkno = 2229 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2230 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2231 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2232 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str, 2233 (unsigned long long)osb->root_blkno, 2234 (unsigned long long)osb->system_dir_blkno, 2235 osb->s_clustersize_bits); 2236 2237 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2238 if (!osb->osb_dlm_debug) { 2239 status = -ENOMEM; 2240 mlog_errno(status); 2241 goto out_uuid_str; 2242 } 2243 2244 atomic_set(&osb->vol_state, VOLUME_INIT); 2245 2246 /* load root, system_dir, and all global system inodes */ 2247 status = ocfs2_init_global_system_inodes(osb); 2248 if (status < 0) { 2249 mlog_errno(status); 2250 goto out_dlm_out; 2251 } 2252 2253 /* 2254 * global bitmap 2255 */ 2256 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2257 OCFS2_INVALID_SLOT); 2258 if (!inode) { 2259 status = -EINVAL; 2260 mlog_errno(status); 2261 goto out_system_inodes; 2262 } 2263 2264 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2265 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters; 2266 iput(inode); 2267 2268 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, 2269 osb->s_feature_incompat) * 8; 2270 2271 status = ocfs2_init_slot_info(osb); 2272 if (status < 0) { 2273 mlog_errno(status); 2274 goto out_system_inodes; 2275 } 2276 2277 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM); 2278 if (!osb->ocfs2_wq) { 2279 status = -ENOMEM; 2280 mlog_errno(status); 2281 goto out_slot_info; 2282 } 2283 2284 return status; 2285 2286 out_slot_info: 2287 ocfs2_free_slot_info(osb); 2288 out_system_inodes: 2289 ocfs2_release_system_inodes(osb); 2290 out_dlm_out: 2291 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2292 out_uuid_str: 2293 kfree(osb->uuid_str); 2294 out_journal: 2295 kfree(osb->journal); 2296 out_orphan_wipes: 2297 kfree(osb->osb_orphan_wipes); 2298 out_slot_recovery_gen: 2299 kfree(osb->slot_recovery_generations); 2300 out_vol_label: 2301 kfree(osb->vol_label); 2302 out_recovery_map: 2303 kfree(osb->recovery_map); 2304 out: 2305 kfree(osb); 2306 sb->s_fs_info = NULL; 2307 return status; 2308 } 2309 2310 /* 2311 * will return: -EAGAIN if it is ok to keep searching for superblocks 2312 * -EINVAL if there is a bad superblock 2313 * 0 on success 2314 */ 2315 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2316 struct buffer_head *bh, 2317 u32 blksz, 2318 struct ocfs2_blockcheck_stats *stats) 2319 { 2320 int status = -EAGAIN; 2321 2322 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2323 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2324 /* We have to do a raw check of the feature here */ 2325 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2326 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2327 status = ocfs2_block_check_validate(bh->b_data, 2328 bh->b_size, 2329 &di->i_check, 2330 stats); 2331 if (status) 2332 goto out; 2333 } 2334 status = -EINVAL; 2335 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 2336 mlog(ML_ERROR, "found superblock with incorrect block " 2337 "size: found %u, should be %u\n", 2338 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 2339 blksz); 2340 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2341 OCFS2_MAJOR_REV_LEVEL || 2342 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2343 OCFS2_MINOR_REV_LEVEL) { 2344 mlog(ML_ERROR, "found superblock with bad version: " 2345 "found %u.%u, should be %u.%u\n", 2346 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2347 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2348 OCFS2_MAJOR_REV_LEVEL, 2349 OCFS2_MINOR_REV_LEVEL); 2350 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2351 mlog(ML_ERROR, "bad block number on superblock: " 2352 "found %llu, should be %llu\n", 2353 (unsigned long long)le64_to_cpu(di->i_blkno), 2354 (unsigned long long)bh->b_blocknr); 2355 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2356 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2357 mlog(ML_ERROR, "bad cluster size found: %u\n", 2358 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2359 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2360 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2361 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2362 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2363 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2364 mlog(ML_ERROR, 2365 "Superblock slots found greater than file system " 2366 "maximum: found %u, max %u\n", 2367 le16_to_cpu(di->id2.i_super.s_max_slots), 2368 OCFS2_MAX_SLOTS); 2369 } else { 2370 /* found it! */ 2371 status = 0; 2372 } 2373 } 2374 2375 out: 2376 if (status && status != -EAGAIN) 2377 mlog_errno(status); 2378 return status; 2379 } 2380 2381 static int ocfs2_check_volume(struct ocfs2_super *osb) 2382 { 2383 int status; 2384 int dirty; 2385 int local; 2386 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2387 * recover 2388 * ourselves. */ 2389 2390 /* Init our journal object. */ 2391 status = ocfs2_journal_init(osb, &dirty); 2392 if (status < 0) { 2393 mlog(ML_ERROR, "Could not initialize journal!\n"); 2394 goto finally; 2395 } 2396 2397 /* Now that journal has been initialized, check to make sure 2398 entire volume is addressable. */ 2399 status = ocfs2_journal_addressable(osb); 2400 if (status) 2401 goto finally; 2402 2403 /* If the journal was unmounted cleanly then we don't want to 2404 * recover anything. Otherwise, journal_load will do that 2405 * dirty work for us :) */ 2406 if (!dirty) { 2407 status = ocfs2_journal_wipe(osb->journal, 0); 2408 if (status < 0) { 2409 mlog_errno(status); 2410 goto finally; 2411 } 2412 } else { 2413 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not " 2414 "unmounted cleanly, recovering it.\n", osb->dev_str); 2415 } 2416 2417 local = ocfs2_mount_local(osb); 2418 2419 /* will play back anything left in the journal. */ 2420 status = ocfs2_journal_load(osb->journal, local, dirty); 2421 if (status < 0) { 2422 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2423 goto finally; 2424 } 2425 2426 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 2427 jbd2_journal_set_features(osb->journal->j_journal, 2428 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2429 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2430 else 2431 jbd2_journal_clear_features(osb->journal->j_journal, 2432 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2433 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2434 2435 if (dirty) { 2436 /* recover my local alloc if we didn't unmount cleanly. */ 2437 status = ocfs2_begin_local_alloc_recovery(osb, 2438 osb->slot_num, 2439 &local_alloc); 2440 if (status < 0) { 2441 mlog_errno(status); 2442 goto finally; 2443 } 2444 /* we complete the recovery process after we've marked 2445 * ourselves as mounted. */ 2446 } 2447 2448 status = ocfs2_load_local_alloc(osb); 2449 if (status < 0) { 2450 mlog_errno(status); 2451 goto finally; 2452 } 2453 2454 if (dirty) { 2455 /* Recovery will be completed after we've mounted the 2456 * rest of the volume. */ 2457 osb->local_alloc_copy = local_alloc; 2458 local_alloc = NULL; 2459 } 2460 2461 /* go through each journal, trylock it and if you get the 2462 * lock, and it's marked as dirty, set the bit in the recover 2463 * map and launch a recovery thread for it. */ 2464 status = ocfs2_mark_dead_nodes(osb); 2465 if (status < 0) { 2466 mlog_errno(status); 2467 goto finally; 2468 } 2469 2470 status = ocfs2_compute_replay_slots(osb); 2471 if (status < 0) 2472 mlog_errno(status); 2473 2474 finally: 2475 kfree(local_alloc); 2476 2477 if (status) 2478 mlog_errno(status); 2479 return status; 2480 } 2481 2482 /* 2483 * The routine gets called from dismount or close whenever a dismount on 2484 * volume is requested and the osb open count becomes 1. 2485 * It will remove the osb from the global list and also free up all the 2486 * initialized resources and fileobject. 2487 */ 2488 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2489 { 2490 /* This function assumes that the caller has the main osb resource */ 2491 2492 /* ocfs2_initializer_super have already created this workqueue */ 2493 if (osb->ocfs2_wq) 2494 destroy_workqueue(osb->ocfs2_wq); 2495 2496 ocfs2_free_slot_info(osb); 2497 2498 kfree(osb->osb_orphan_wipes); 2499 kfree(osb->slot_recovery_generations); 2500 /* FIXME 2501 * This belongs in journal shutdown, but because we have to 2502 * allocate osb->journal at the middle of ocfs2_initialize_super(), 2503 * we free it here. 2504 */ 2505 kfree(osb->journal); 2506 kfree(osb->local_alloc_copy); 2507 kfree(osb->uuid_str); 2508 kfree(osb->vol_label); 2509 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2510 memset(osb, 0, sizeof(struct ocfs2_super)); 2511 } 2512 2513 /* Depending on the mount option passed, perform one of the following: 2514 * Put OCFS2 into a readonly state (default) 2515 * Return EIO so that only the process errs 2516 * Fix the error as if fsck.ocfs2 -y 2517 * panic 2518 */ 2519 static int ocfs2_handle_error(struct super_block *sb) 2520 { 2521 struct ocfs2_super *osb = OCFS2_SB(sb); 2522 int rv = 0; 2523 2524 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2525 pr_crit("On-disk corruption discovered. " 2526 "Please run fsck.ocfs2 once the filesystem is unmounted.\n"); 2527 2528 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) { 2529 panic("OCFS2: (device %s): panic forced after error\n", 2530 sb->s_id); 2531 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) { 2532 pr_crit("OCFS2: Returning error to the calling process.\n"); 2533 rv = -EIO; 2534 } else { /* default option */ 2535 rv = -EROFS; 2536 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb))) 2537 return rv; 2538 2539 pr_crit("OCFS2: File system is now read-only.\n"); 2540 sb->s_flags |= SB_RDONLY; 2541 ocfs2_set_ro_flag(osb, 0); 2542 } 2543 2544 return rv; 2545 } 2546 2547 int __ocfs2_error(struct super_block *sb, const char *function, 2548 const char *fmt, ...) 2549 { 2550 struct va_format vaf; 2551 va_list args; 2552 2553 va_start(args, fmt); 2554 vaf.fmt = fmt; 2555 vaf.va = &args; 2556 2557 /* Not using mlog here because we want to show the actual 2558 * function the error came from. */ 2559 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV", 2560 sb->s_id, function, &vaf); 2561 2562 va_end(args); 2563 2564 return ocfs2_handle_error(sb); 2565 } 2566 2567 /* Handle critical errors. This is intentionally more drastic than 2568 * ocfs2_handle_error, so we only use for things like journal errors, 2569 * etc. */ 2570 void __ocfs2_abort(struct super_block *sb, const char *function, 2571 const char *fmt, ...) 2572 { 2573 struct va_format vaf; 2574 va_list args; 2575 2576 va_start(args, fmt); 2577 2578 vaf.fmt = fmt; 2579 vaf.va = &args; 2580 2581 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV", 2582 sb->s_id, function, &vaf); 2583 2584 va_end(args); 2585 2586 /* We don't have the cluster support yet to go straight to 2587 * hard readonly in here. Until then, we want to keep 2588 * ocfs2_abort() so that we can at least mark critical 2589 * errors. 2590 * 2591 * TODO: This should abort the journal and alert other nodes 2592 * that our slot needs recovery. */ 2593 2594 /* Force a panic(). This stinks, but it's better than letting 2595 * things continue without having a proper hard readonly 2596 * here. */ 2597 if (!ocfs2_mount_local(OCFS2_SB(sb))) 2598 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2599 ocfs2_handle_error(sb); 2600 } 2601 2602 /* 2603 * Void signal blockers, because in-kernel sigprocmask() only fails 2604 * when SIG_* is wrong. 2605 */ 2606 void ocfs2_block_signals(sigset_t *oldset) 2607 { 2608 int rc; 2609 sigset_t blocked; 2610 2611 sigfillset(&blocked); 2612 rc = sigprocmask(SIG_BLOCK, &blocked, oldset); 2613 BUG_ON(rc); 2614 } 2615 2616 void ocfs2_unblock_signals(sigset_t *oldset) 2617 { 2618 int rc = sigprocmask(SIG_SETMASK, oldset, NULL); 2619 BUG_ON(rc); 2620 } 2621 2622 module_init(ocfs2_init); 2623 module_exit(ocfs2_exit); 2624